Multiple Hosting Accounts for Redundancy

My client is running an ecommerce store with a pretty basic server configuration. As such, we can get hosting pretty much anywhere to fit our needs for about $50/month. However, as it’s an ecommerce store she’s very worried about redundancy. I understand that any decent host will have a SLA but that doesn’t really help us because we’re not worried about being credited back the cost of hosting, we just don’t want the site to go down in the first place.

If I were to mirror the same website/database on multiple servers with different hosting companies and add both sets of nameservers to the domain name, would that act as a load balancer and be redundant in case of an outage? Granted I would need to deal with how to merge the database data together on a periodic basis, it’s possible. My concern is that if the web server goes down on hosting company A, the nameservers are still going to be online which means that the site will still be down.

Not trying to say that this is all a good idea, just attempting to explore my options.

What you’ve suggested won’t work. The easiest way to achieve what you want is failover DNS and set up your database for replication, however it has limitations with regards to TTL and how long it takes for visitors to be redirected to the backup server as even with a short TTL set, some ISPs will ignore it and have their own cache time, so there won’t be an instantaneous failover.

There are other ways you can set up better redundancy using e.g haproxy and keepalived but it can be complex to set up.

You should probably analyse what the real cost would actually be to the client of a period of downtime and then compare to the set up and ongoing costs of a high availability hosting configuration - I would suggest that a client with the level of budget for hosting you’ve mentioned is in reality unlikely to truly need bulletproof failover.

Hi,

What you’ve suggested will probably result in more downtime than just an account with a reputable hosting company. Unfortunately high availability rarely comes cheaply and what has already been suggested is a good idea - sit down and work out how much an hour of downtime would cost your client and go from there. If you’re coming up with anything less than $500 then I’d say it’s not worth it.

Thanks,

Sounds great - thanks for the help. $500 seems reasonable if we can keep it around there.

$500 won’t get you into the room to talk to someone who understands high availability never-mind something that can handle it. You’d be much better off spending that money on a high-quality host who has relatively redundant systems such as multiple paths to the internet and a big old generator with a bigger tank of diesel fuel.

I actually said if you’re not losing at least $500/hr then it’s most likely not going to be worth it :slight_smile: Not that $500 was a starting budget :slight_smile: